[Fix] Fix logprob and normalized_logprob (#1428)

python/sglang/lang/backend/runtime_endpoint.py

@@ -239,9 +239,12 @@ class RuntimeEndpoint(BaseBackend):
         # Compute logprob
         data = {
             "text": [s.text_ + c for c in choices],
-            "sampling_params": {"max_new_tokens": 0},
+            "sampling_params": {
+                "max_new_tokens": 0,
+                "temperature": 0,
+            },
             "return_logprob": True,
-            "logprob_start_len": max(prompt_len - 2, 0),
+            "logprob_start_len": max(prompt_len - 2, 0),  # for token healing
         }
         obj = self._generate_http_request(s, data)
 

python/sglang/lang/interpreter.py

@@ -9,7 +9,7 @@ import uuid
 import warnings
 from concurrent.futures import ThreadPoolExecutor
 from contextlib import contextmanager
-from typing import Any, Callable, Dict, List, Optional, Union
+from typing import Any, Callable, Dict, List, Optional
 
 import tqdm